Skip to content

AI visibility measurement guide · By Yes AI

How long do AI visibility changes take to show?

Published 5 min read

Owners reasonably want to know when a change will show up in AI answers. Some steps have published delays measured in hours or days, while others depend on recrawls and model updates that no operator schedules around you. This guide lists what is documented and what is not.

Quick answer

Some delays are documented. OpenAI says robots.txt changes take about 24 hours to reach its search systems, Perplexity says up to 24 hours, and Google says a Search generative AI control change generally takes a few days. Google says a recrawl can take a few days to a few weeks. Parts of an answer drawn from a model's training data change only when the operator updates the model, so results are measured over months.

Office manager in a Townsville office marking review dates on a paper wall planner beside a laptop

Which changes take effect within days?

Changes to crawler access have the shortest documented delays. OpenAI says that, for search results, its systems can take about 24 hours to adjust after a robots.txt update. Perplexity says changes may take up to 24 hours to be reflected. Google says a change to the Search generative AI control in Search Console generally takes a few days to exclude a site's content from AI features, with some content taking longer because of caching. These delays cover permission only. They tell you when a crawler may visit or when an exclusion applies, and nothing about when a page will be cited.

How long does Google take to see a page change?

Google's recrawl documentation says crawling can take anywhere from a few days to a few weeks, and that requesting a crawl does not mean the page will be included in results instantly or at all. An owner or full user of the Search Console property can request a recrawl of individual pages with the URL Inspection tool, and a sitemap helps for many pages. Google says there is a quota for individual requests and that asking repeatedly for the same URL does not speed it up. In its AI features guidance, Google says crawling can take from several days to several months, depending on how often its systems decide a page needs refreshing.

Why do some AI answers lag behind your website?

An assistant can answer from its model's training data, from a live search, or from both. The part that comes from training data reflects the web at the time the model was trained, and it changes only when the operator releases an updated model. Operators do not publish a schedule a business can plan around. The part that comes from a live search can pick up your corrected page once it has been crawled and indexed, but it can also pick up an older copy of the detail on another site. That is why a fix can show in one assistant while another still repeats the old version.

Why do results move from week to week?

Answers are generated afresh for each question, and the same question can produce different wording and different businesses on different runs. Google says AI Mode and AI Overviews may use different models and techniques, so the links they show will vary, and that AI Overviews appear only when its systems judge them useful. The search results feeding the answers also change as other sites publish and update pages. A weekly movement in a handful of prompt checks is therefore weak evidence. Look at the same questions over several months and judge the pattern.

What is a sensible measurement schedule?

Set the schedule before work starts and keep it regardless of early results, so nobody is tempted to measure only on good days. The baseline comes first: the full set of buyer questions, run and saved before any change. Crawler access fixes can be confirmed in server logs within days. Page changes need Google's recrawl time before any effect can be expected. Prompt checks then run at a fixed interval with the same questions and settings.

Hypothetical example: a business fixes a CDN rule blocking OAI-SearchBot and rewrites four service pages. It saves a baseline of 30 questions before the work, checks its logs a few days after the CDN change, confirms in Search Console a few weeks later that the edited pages have been recrawled, and reruns the 30 questions once a month for six months before judging the result. These numbers are illustrative assumptions.

What should a provider tell you about timing?

A provider should separate documented delays, such as robots.txt processing, from estimates, such as when answers might change. Be cautious of anyone who promises your business will be named within a fixed number of weeks, because no operator publishes when its models will next be updated or how its answers choose businesses. Ask instead when each piece of work will be delivered, when each measurement will be taken and what will be compared. For Yes AI's own work, timing is confirmed in your written quote after a short scoping conversation.

Documented delays for common AI visibility changes

ChangeDocumented delaySourceWhat it does not tell you
robots.txt change for OpenAI crawlersAbout 24 hours for search systems to adjustOpenAI crawler documentationWhether ChatGPT will cite the page
robots.txt change for Perplexity crawlersUp to 24 hoursPerplexity crawler documentationWhether Perplexity will name you
Search generative AI control set to ExcludeGenerally a few daysGoogle Search Console HelpAnything about sites left at Include
Page edit with a recrawl requestA few days to a few weeksGoogle Search CentralWhether the page is used in an AI feature
Preview control such as nosnippetSeveral days to several monthsGoogle Search CentralThe date a given page will be recrawled
Model training dataNo published scheduleNone publishedAny date to plan around

Practical checklist

  • A baseline prompt set has been run and saved before any change.
  • Each change is logged with its date so later observations can be compared.
  • Crawler access fixes are confirmed in server logs a few days after release.
  • Google indexing of edited pages is checked in the URL Inspection tool.
  • Re-check dates are set in advance and kept, whatever the early results.
  • Any timing in a proposal is stated as an estimate and confirmed in the written quote.

Common questions

Can we speed up Google's recrawl?

Partly. An owner or full user of the Search Console property can request a recrawl of individual pages with the URL Inspection tool, and a sitemap helps Google discover many changed pages. Google says there is a quota, that repeated requests for the same URL do not get it crawled faster, and that a crawl request does not mean the page will be included.

Why does an assistant still show our old address after we fixed the website?

It may be answering from training data gathered before the fix, which changes only when the model is updated. Or its search may have found another page that still carries the old address, such as a directory listing. Check the cited links in the answer, correct any third-party pages you find, and check again later with the same question.

How often should we rerun prompt checks?

Pick a fixed interval and keep to it. Monthly is a practical default for a small business, because it gives recrawls time to happen and builds enough runs to see a pattern over a few months. Use the same questions, assistants and settings each time, and save every result, including the ones that look bad.

Is three months long enough to judge AI visibility work?

It depends on what changed. Crawler access fixes can be confirmed within days. Effects that rely on Google recrawling pages need weeks, and anything that relies on a model update has no published timeline. Agree review points in advance, compare the same questions under the same conditions, and treat early results as provisional.

Do paid monitoring tools show changes faster?

They can run prompts more often, so you see movements sooner. They cannot shorten the underlying delays: crawlers still need to revisit, search indexes still need to update and models still change on the operator's schedule. More frequent checks also show more run-to-run variation, so read them as trends.

Sources and further reading

Continue planning

Plan an evidence-led AI visibility review

Describe your business category, service area and the questions buyers ask. We can discuss a repeatable visibility baseline and which public content needs review. Keep customer records and confidential information out of this enquiry.